Single pixel imaging at high pixel resolutions.

Area of Science:

• Optics and Photonics
• Computational Imaging
• Signal Processing

Background:

• Single pixel imaging (SPI) typically yields low resolutions (32x32 to 256x256) due to trade-offs in compression, modulation frequency, and reconstruction time.
• Existing SPI research has not significantly improved resolution over the past decade.
• Classical imaging methods offer higher resolutions but lack the advantages of SPI.

Purpose of the Study:

• To demonstrate high-resolution image acquisition using single pixel imaging (SPI) at the full resolution of the digital micromirror device (DMD).
• To develop novel sampling and reconstruction strategies for enabling real-time, high-resolution SPI.
• To overcome the limitations of current SPI techniques in terms of speed and resolution.

Main Methods:

• Proposed differential, binary, non-adaptive sampling strategies that incorporate image partitioning for field-of-view determination.
• Developed a reconstruction algorithm based on differential Fourier domain regularized inversion (D-FDRI).
• Achieved image measurement at the full DMD resolution (1024x768) for sparse images within 0.3 seconds.

Main Results:

• Successfully reconstructed sparse images at 1024x768 resolution in 0.3 seconds.
• Achieved a compression ratio of approximately 0.4%, enabling an acquisition frequency of 7Hz.
• Demonstrated that non-sparse images can be reconstructed with reduced detail.

Conclusions:

• The proposed SPI framework offers a viable alternative to adaptive SPI and classical compressive sensing methods for high-resolution imaging.
• This approach enables real-time, high-resolution imaging previously unattainable with SPI.
• The method is particularly effective for sparse images or situations with an unknown, limited field of view.